Fusion of ERP and fNIRS signals for classification of internet gaming disorder in an attention-modulating dot-probe task

Background and Objective Internet gaming disorder (IGD) has emerged as a significant behavioral addiction and public health concern. However, its underlying neurobiological mechanisms remain unclear. This study aimed to develop a machine learning approach to evaluate IGD based on multimodal neurophysiological signals. Methods Forty-seven male university students with IGD and 42 recreational game users (RGUs) participated in the study. Event-related potential (ERP) and functional near-infrared spectroscopy (fNIRS) signals were simultaneously recorded during a game-related dot-probe paradigm to evaluate attentional bias and cue-reactivity. The extracted features included general linear model (GLM) beta values representing changes in cortical oxygenation from fNIRS, along with the latency and amplitude of ERP components. Results Compared to RGUs, IGD participants exhibited altered P100, N200, and P300 responses that were modulated by stimulus duration. Additionally, they showed reduced prefrontal hemodynamic activity, particularly in the dorsolateral prefrontal cortex. Several machine learning algorithms were evaluated for classification, including baseline classifiers, ensemble methods, and a hybrid model, with the KNN algorithm yielding the highest classification performance. Classification accuracy increased from 79.95% using ERP features alone to 85.4% when combining ERP and fNIRS features. Conclusions The findings suggest that IGD and RGU can be objectively distinguished using neurophysiological data, offering a potential tool to support clinical diagnosis. Reduced activation in the dorsolateral prefrontal cortex may serve as a potential neuro-biomarker for IGD. In addition, altered N200 and P300 responses, which are modulated by cue duration, may also serve as electrophysiological markers of IGD. This is the first study to classify IGD using multimodal neural responses during the dot-probe task. This comprehensive approach, combining electrophysiological and hemodynamic data within a cue-reactivity framework, highlights the potential of multimodal imaging to enhance our understanding and detection of IGD.